Industrializing Physics-Driven Generative Design
Interview with Thomas Rees of ToffeeX
Introduction to ToffeeX
Can you give us a brief overview of ToffeeX and the company and your software capabilities?
We are a software company based in London with a mission to change how engineers approach and use physics-driven generative design techniques, such as topology optimization. Specifically, we specialize in generative design for thermo-fluid engineering components.
Maybe the best way to understand our company, our product, and its capabilities is to understand the story behind our name. We started in 2020 as TOffeeAM, which stands for Topology Optimization for fluids engineering (TOffee) and Additive Manufacturing (AM).
During those first few years, we proved the effectiveness of our product at creating high-performance designs for thermo-fluids components manufactured using additive processes. Some examples of the successes we achieved with our customers included radical new designs for additively manufactured natural gas vaporizers, two-fluid heat exchangers, and cold plates for power electronics.
However, through our interactions with our customers we quickly realized a broader market need: the ability to manufacture these optimized designs not just with AM, but also with established, low-cost, mass-production techniques such as machining, stamping, and so on. This sparked approximately a year of intensive development, resulting in a comprehensive set of features which allow our customers to create designs which can be manufactured not just with AM, but also conventional approaches. The X in our new name reflects this expanded capability.
From a business perspective, most of our customers are large engineering corporations such as Baker Hughes, Rolls-Royce, Toyota, and so on. Our largest customer base is in Japan, where we have a collaboration agreement with SCSK who handle sales, support, and training in that market. We have raised approximately £6M in VC funding and have a team of approximately 30 people based in central London, UK.
CDFAM Presentation Focus
What will be the main focus of your presentation at the upcoming CDFAM event in Berlin?
Our goal at CDFAM is to continue to shift perceptions surrounding physics-driven generative techniques in engineering.
Just last week I was working our stand at a trade show, and a regular comment I received from visitors could essentially be boiled down to ‘OK sure, but no-one can actually manufacture components generated with TO, even with Additive’. This is a very common (and deserved) skepticism.
From our inception, our vision at ToffeeX has been to empower engineers to create functionally superior component designs, not just pretty pictures which look great in a slide deck.
As I described earlier, I think our two key technologies which enable this are our manufacturability constraints and our complex multi-physics simulation models.
From the design of thermofluids components like heat exchangers, to structural considerations such as pressure containment and thermal stress minimization, physics-driven generative design tools can now offer engineers tangible solutions.
To solidify this message, at CDFAM we will share success stories from our customers. From groundbreaking applications of printed parts to precision-machined components, we think real examples speak volumes.
Overcoming Manufacturing Misconceptions
How does ToffeeX assist potential clients in overcoming the misconception that topology optimization produces parts that cannot be manufactured?
The answer to this question is quite simple, and there are no two ways about it – the proof is in the proverbial pudding. The easiest way we’ve found to change the minds of prospective clients is by showing them proven successes, like the heat sink below, a design created for our customer Ricoh, who manufactured it with their binder-jetting additive process, before analyzing its real-world performance. Presented with this sort of evidence, we normally see attitudes change quickly!
In cases where we don’t have any results we can publicly share, our customer success team steps in. They develop proof-of-concept designs, allowing our clients to validate our capabilities using their own simulation, analysis, and manufacturing tools. Ultimately, it is all about showing candidates the value we can bring first-hand.
On a personal note, seeing manufactured and tested components is what gets me out of bed in the morning. It is my primary motivation. With a background in experimental aerodynamics I’ve always been more motivated by hardware and ‘real-life’ than by simulation and computation. Despite the time it can take from a project kick-off to the final manufacturing and testing of a component, the moment I can touch a real part and see its performance data gives me a thrill like absolutely no other. It makes it all worth it.
We’ve got a few more success stories in the pipeline, such as some intricate two-fluid heat exchanger designs which have been manufactured and tested by aerospace clients and show promising results. Stay tuned for our presentations at ASME TurboExpo 2024, where we’ll present these results.
Can you discuss some of the performance objectives, and manufacturing constraints that can be defined in your software?
Absolutely. Maybe it is first worth taking some time to first briefly describe the ToffeeX user experience, to place these features in their context.
We offer a workflow based around the concept of a toolbox – this is a collection of different modules representing the different aspects of a ToffeeX run: the design domain, the design or optimization problem, a description of the physics or simulation problem, and finally a meshing tool.
This modular workflow allows users to mix and match different modules within a ToffeeX Toolbox to create and recreate unique designs and simulations, for example by keeping the design requirements constant while exploring different simulation problems (such as different operating conditions).
Now, with this context, it’s no surprise that the Design module is where the magic primarily happens! It allows users to specify multi-objective optimization problems, combining various design objectives weighted according to the user’s requirements. These design objectives may include the minimization of total pressure losses across the system, or minimizing the temperature in a region of the domain.
Alongside the objective function terms, a wide range of design constraints is available, which ensure that critical design requirements like weight, manufacturability, or structural integrity are met. Examples include aspects like weight, minimum feature size, bit size (for machining) and so on.
To make this a little more concrete, I’ll provide an example which shows the effect of one of the additive manufacturing constraints available in ToffeeX: the ability to control the cross-sectional area of a part along a chosen axis.
This is useful for our customers choosing to build with Additive and who want to reduce the effects of warpage or shrink lines along their build. In this image, the unconstrained design generated by ToffeeX is on the left, while the constrained design is shown on the right. The plots show the cross sectional area along the axis. I think the effect speaks for itself!
All these features are accessible through a user-friendly interface. We’ve invested heavily in UI/UX refinement, aiming for simplicity. We ended up settling on a straightforward click-and-set approach, where users can configure settings effortlessly (the image below shows all the data a user needs to know to set up an overhang reduction constraint – you can see that it is not a long list of parameters!).
Real-World Applications
Could you provide some examples of how your solutions have been applied in real-world scenarios?
I’m thinking back to our earlier discussion on how we convince prospective clients of the value proposition of physics-driven generative design tools… I suspect there is going to be some overlap in my answers, but let me try to provide some real-world scenarios where our solutions have made significant impacts.
One instance involved a natural gas vaporizer designed for a customer using ToffeeX and manufactured with Laser Powder Bed Fusion. Upon testing, the performance exceeded expectations, surpassing the operational limits of the test circuit. This necessitated a redesign to reduce the heat transfer performance to within acceptable bounds.
We’ve also undertaken multiple projects to re-design two-fluid heat exchangers for clients in power systems and aerospace. These designs accommodated various working fluids such as air, nitrogen, water, and mineral oil. Additionally, structural considerations like pressure containment and external loading on the heat exchangers were addressed, alongside factors like the reduction of the size and weight. Across the board, our designs met or exceeded performance requirements post-manufacture and testing.
In another field, we’ve engineered cooling systems for a variety of turbomachinery applications, including lubricated bearings. Experimental testing confirmed the effectiveness of these components in reducing system temperatures and therefore extending component longevity.
Finally, our software has even found application in more “exotic” fields like the design of adsorption contactors for gas separation, with experimental results indicating a significant reduction in the power required to drive fluid through the adsorption contactor.
It is probably fair to say that these examples demonstrate the breadth of applications where ToffeeX has been successfully used. It is true that our customer base is currently primarily concentrated in aerospace and energy systems industries, where our solutions have demonstrated significant value to our customers. However, we’re witnessing rapid growth in markets associated with the energy transition, including hydrogen technologies, thermal management for power electronics, and emerging fields like carbon capture and storage.
Takeaways and Goals for CDFAM
What are the key takeaways you hope attendees will leave with from your presentation at CDFAM, and what do you hope to gain from participating in the event?
Other than hoping other attendees will be convinced by our vision of physics-driven generative design…? I’m also hoping that the success stories I’ll be sharing will be catalysts for collaboration and connection – physics driven tools like topology optimization are only a small part of the computational design space and they inevitably need to be combined with other technologies to maximize their impact.
As for my personal objectives, I’m just primarily interested in hearing from all the other speakers, not only during their presentations, but also over a cup/mug/bucket of coffee! I’ve only been working in the generative design space for about 4 years now and I still feel I have so much to learn.
